Method and system for evaluating functional health index of old people

Abstract

The invention provides a functional health index assessment method and system for old people, and the method comprises the steps: obtaining health influence factor tracking survey queue data of old people, and building a feature set through the combination of a domain knowledge base; modeling a mapping relationship between the feature set and the label by using a machine learning model, and constructing a total loss function to train the machine learning model to obtain the trained machine learning model, and screening out functional health key factors to construct a new important feature set; and establishing multiple logistic regression based on the new important feature set, outputting the advantage ratio of each important feature, converting the advantage ratio into a questionnaire score, defining a scoring rule, designing a questionnaire, setting a questionnaire total score screening threshold, and dividing risk crowds. According to the method, multi-dimensional features are integrated through machine learning, functional health key factors of the aged are screened and converted into simple questionnaires, rapid community evaluation of functional health of the aged is achieved, and scientificity and universality are both achieved.

Description

Technical Field

[0001] This invention relates to the field of medical and health information technology, and in particular to a method and system for assessing the functional health index of the elderly population. Background Technology

[0002] Population aging has become a global social problem, with my country's aging population ratio significantly higher than the global average, indicating that Chinese society has entered a phase of deep aging. As the aging process accelerates, the caregiving pressure on families and society is increasing daily, and the expenditure burden on the social security system continues to rise. Therefore, addressing the supply and demand imbalance in elderly care and health services, and improving the overall level of health management and services for the elderly, are crucial issues in the current healthcare field.

[0003] Focusing on the health of the elderly requires not only exploring treatments for diseases, but more importantly, assessing their functional health status before diseases occur. Accurately assessing the functional health status of the elderly and subsequently classifying them into risk levels will significantly promote the improvement of health services for the elderly. The community is the primary place for the elderly to address their medical needs, and health screenings for the elderly are usually conducted alongside annual physical examinations. Due to limitations in medical equipment and human resources, some complex and time-consuming "gold standard" assessment tools that rely on specific medical equipment cannot be used in the community. Therefore, designing a fast and accurate functional health assessment tool for the elderly, suitable for daily health management in community settings, has significant practical significance and application value.

[0004] Currently, in the field of functional health status assessment of the elderly, the existing research and technical solutions have the following problems: (1) Most health evaluation indicators are based on the health data of the entire adult population and have not been further optimized and adjusted for the elderly population, resulting in poor applicability when actually applied to the elderly population. For example, the invention patent with publication number CN119538060A discloses a health risk assessment system based on the fusion analysis of multi-source data from smart wearable devices. Based on the data-driven approach, it fuses multi-source data, dynamically adjusts the assessment weights based on the status, and outputs a comprehensive coefficient to assess the health level. However, this model is based on the data of the entire adult population and has not been optimized for the characteristics of the elderly. It may lead to assessment bias due to the general threshold, resulting in poor applicability among the elderly population.

[0005] (2) There is a lack of multi-dimensional health assessments for the elderly from the perspective of functional health. For example, the invention patent with publication number CN118760850A discloses a digital screening and remote detection method for comprehensive assessment of the elderly. After acquiring the vital signs data of the elderly through sensors, different weighted values ​​are calculated for different elderly people using an elderly health assessment model to constitute the elderly health score. This invention conducts individualized health detection for the elderly from multiple dimensions, but its data acquisition module relies on wearable devices, and the assessment only relies on the raw sensor data (heart rate, acceleration), which is not converted into functional health indicators and remains at the signal processing level, lacking an assessment from the perspective of functional health.

[0006] (3) Cumbersome and time-consuming multi-dimensional assessment tools are difficult to promote and use in community health check-up scenarios. For example, the invention patent with publication number CN118629659A discloses a method and system for monitoring and assessing the health of the elderly based on big data. By acquiring the physical data and types of chronic diseases of the elderly, a highly accurate health assessment model for the elderly is constructed based on big data, which can obtain the disease risk index of the elderly. However, this invention relies heavily on the continuous collection of large-scale historical data and cloud modeling. It requires wearable terminals, a complete database, complex curve fitting and risk calculation processes. The assessment process involves multiple steps and a large amount of calculation, which is not conducive to rapid and lightweight on-site screening and is difficult to promote and use in community scenarios.

[0007] In summary, focusing on the field of functional health status assessment for the elderly, improving the applicability of health assessment status in the elderly population, achieving rapid and accurate functional health status assessment of the elderly through multi-dimensional health, and promoting its application in communities are urgent issues that need to be addressed. Summary of the Invention

[0008] The purpose of this invention is to provide a method and system for assessing the functional health index of the elderly. This assessment method focuses on the elderly in community settings and is used to solve the problems of poor applicability and implementation difficulties of existing health assessment methods for the elderly in community settings, as well as the lack of multi-dimensional health assessment methods for health functions. It provides a simple and efficient tool for early screening of the functional health status of the elderly.

[0009] An embodiment of the method for assessing the functional health index of the elderly includes the following steps: Step 1: Obtain cohort data from the long-term study on health influencing factors in the elderly population, combine it with the knowledge base in the field of health assessment and management of the elderly, determine the effective features, and preprocess the effective features to obtain the feature set; Step 2: Use a machine learning model to model the mapping relationship between the preprocessed feature set and the labels, and construct a total loss function to train the machine learning model to obtain the trained machine learning model. Select key factors of functional health to construct a new set of important features. Step 3: Based on the new set of important features, establish a multivariate logistic regression to output the advantage ratio of each important feature. Convert the advantage ratio of each important feature into a questionnaire score and define the scoring rules. Design a functional health index assessment questionnaire for the elderly population, set a total score screening threshold for the questionnaire, and classify the risk population.

[0010] In one embodiment, step 1, preprocessing the effective features to obtain a feature set, includes: converting the effective features into the original dataset, and performing missing value imputation and variable type conversion on the original dataset to obtain the feature set; The missing value imputation includes: removing samples with a missing value ratio exceeding 50% from the original dataset; and for the missing values ​​in the retained samples, using the K-nearest neighbor algorithm for imputation, calculated as follows: , In the formula, Indicates the first number after filling The first sample 1 eigenvalue, This represents the number of neighbors in the K-nearest neighbors algorithm. Representation and Sample The closest in Euclidean distance A set of sample indexes, express Neighbor samples in Indicates the first The first sample There are 1 eigenvalues; where the Euclidean distance is calculated using the following formula: , In the formula, Indicates sample and samples The Euclidean distance between them Indicates the first Feature values ​​of each sample Indicates the first Feature values ​​of each sample For the sample size, The number of sample features; The variable type conversion includes: for continuous variables in the original dataset, calculating the Youden index using ROC curves. Determining the optimal cutoff point for continuous variables And binarize the continuous variables to output the feature set. ,in , In the formula, Indicates the first The first sample Binarized eigenvalues, Indicates the first The first sample The original feature values ​​of each feature. Indicates the first The optimal cutoff point for each feature Indicates the number of samples. The number of sample features.

[0011] Furthermore, continuous variables include: age, height, weight, and waist circumference.

[0012] In one embodiment, in step 2, when modeling the mapping relationship between the preprocessed feature set and the label using a machine learning model, a one-hot encoding strategy and a target encoding strategy are used to encode the preprocessed feature set. Among them, discrete features with multiple values ​​are encoded using a target encoding strategy, which maps the categorical variables of discrete features to continuous variables according to their statistical relationship with the label. The discrete features include dietary habits, education level, and sleep quality.

[0013] In one embodiment, in step 2, the machine learning model is the XGBoost model.

[0014] In one embodiment, in step 2, the constructed total loss function includes a loss term and a regularization term, calculated as follows: , , In the formula, Represents the total loss function. Indicates the loss item. Indicates the first Health outcome labels for each sample Indicates the first Predicted health outcomes for a sample Indicates the number of samples. Represents the regularization term, Indicates the first There are 10 decision trees, where M represents the total number of trees. The regularization term representing the number of single trees, Indicates the leaf splitting threshold. Represents the regularization coefficient. Indicates the first The weight of each leaf node.

[0015] In one embodiment, step 2, which uses cross-validation and grid search to train the machine learning model, specifically includes: dividing the preprocessed feature set into a training set and a test set, using five-fold cross-validation, evaluating the performance of the machine learning model with the F1 score as the evaluation metric, and performing grid search on the hyperparameter space of the machine learning model based on the F1 score as the optimization objective to search for optimized hyperparameters.

[0016] Furthermore, the hyperparameters for search optimization include: learning rate, maximum tree depth, subsampling ratio, feature sampling ratio, and regularization parameter.

[0017] In one embodiment, step 2, which involves selecting key factors for functional health to construct a new set of important features, includes: based on the preprocessed set of features, introducing XGBoost importance weights in the minimum redundancy maximum relevance algorithm framework to perform feature selection, thereby obtaining a new set of important features including age, gender, frequency of consuming white sugar or candy, whether one knows the zodiac animal of the year, whether one can visit neighbors alone, and whether one can lift a 5 kg weight.

[0018] Furthermore, the introduction of XGBoost importance weights for feature selection within the minimum redundancy maximum relevance algorithm framework includes: Based on the preprocessed feature set and the importance of the output parameters of the constructed machine learning model, the mutual information between the features in the feature set and the target variable is calculated and multiplied by the normalized importance weight to evaluate the correlation between each feature and the target variable. Initialize feature subset Select the feature with the highest mutual information. As a subset of features The first feature is selected, and the feature with the highest weighted mutual information is removed from the preprocessed feature set; For the remaining features, calculate their maximum weighted correlation with the target variable and their correlation with the selected feature subset. The difference in minimum redundancy is used as the candidate feature score. The feature with the highest score is selected and removed from the preprocessed feature set. Iteratively select features until a subset of features is reached. Includes predefined By identifying key factors for functional health, a new set of important features is constructed.

[0019] In one embodiment, the multivariate logistic regression calculation formula established in step 3 is as follows: , In the formula, Indicates the number of important features. Indicates the first The first sample One important eigenvalue, Indicates the regression intercept. Represents the regression coefficient. Indicates that under a given feature Under the conditions, the first The health outcome for one sample was death ( The probability of ), Indicates the health outcome label; outputs the odds ratio of each important feature based on the established multivariate logistic regression. .

[0020] In one embodiment, step 3, which involves converting the advantage ratios of each important feature into questionnaire scores and defining scoring rules, includes: For important features with an odds ratio greater than 1, the score for a yes answer is the odds ratio, and the score for a no answer is 1. For important features where the odds ratio is less than 1, the score for a yes answer is 1 / odds ratio, and the score for a no answer is 1. Based on the scoring rules defined above, the calculation formula is as follows: , The total score of the questionnaire is ,in, Indicates the first The advantage ratio of each feature Indicates the first The sample at the th Risk score on each feature Indicates the first The first sample One important eigenvalue.

[0021] In one embodiment, step 3, which involves setting a total questionnaire score screening threshold to classify risk groups, includes: calculating the Youden coefficient, and classifying individuals whose total questionnaire score is greater than or equal to the Youden coefficient as high-risk individuals for functional health within the questionnaire score range.

[0022] On the other hand, the present invention also provides a functional health index assessment system for the elderly, wherein the method for assessing the functional health index of the elderly in the functional health index assessment system includes: The feature selection and preprocessing module is used to obtain cohort data from the long-term survey of health influencing factors of the elderly population, combine it with the knowledge base in the field of health assessment and management of the elderly, establish initial features, and obtain a feature set based on the initial features and preprocessing. The functional health status assessment model construction module is used to model the mapping relationship between the preprocessed feature set and the label using a machine learning model, and to construct a total loss function to train the machine learning model, thereby obtaining the trained machine learning model and screening out key factors of functional health to construct a new set of important features. The Functional Health Status Assessment and Questionnaire Design Module is used to establish a multivariate logistic regression based on a new set of important features to output the advantage ratio of each important feature, convert the advantage ratio of each important feature into a questionnaire score and define the scoring rules, design a functional health index assessment questionnaire for the elderly, set a total score screening threshold for the questionnaire, and classify high-risk groups.

[0023] Compared with the prior art, the beneficial effects of the present invention include at least the following: (1) This invention targets the elderly population and establishes a feature set that comprehensively describes the functional health status of the elderly from the perspective of functional health. Based on the correlation between each feature and health outcome indicators, the seven features with the highest correlation with the functional health status of the elderly are selected, which greatly reduces the dimensionality of the functional health status assessment model of the elderly population and also provides the possibility for its application in the community.

[0024] (2) Using a feature set that comprehensively describes the functional health status of the elderly as input, a basic machine learning model is trained to construct a functional health index assessment model that is adapted to the elderly, which significantly improves the accuracy of prediction and assessment of the functional health status of the elderly and provides a simple and efficient method for early screening of the functional health status of the elderly.

[0025] (3) Based on the constructed functional health index assessment model, we established a multivariate logistic regression model, obtained the parameter OR value to interpret the model, and designed a simple, fast and efficient functional health questionnaire. We set the corresponding questionnaire score thresholds and divided the risk population by the total questionnaire score. This solved the problem of the time and effort required to use complex assessment scales and tools in the community, and is suitable for promotion and use in community scenarios. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0027] Figure 1 This is a flowchart illustrating the functional health index assessment method for the elderly population provided by the present invention.

[0028] Figure 2 The advantage ratio of important features related to the functional health status of the elderly population provided in the embodiments.

[0029] Figure 3This is a schematic diagram of the structure of the functional health index assessment system for the elderly population provided by the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and given in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0031] This invention provides a method for assessing the functional health index of the elderly population, such as... Figure 1 As shown, the method includes the following steps: Step 1: Obtain cohort data from the long-term study on health influencing factors in the elderly population, combine it with the knowledge base of health assessment and management of the elderly, identify effective features, and preprocess the effective features to obtain a feature set.

[0032] Step 1-1: In this embodiment, large-scale cohort data from the long-term study of health influencing factors in the elderly population are obtained. Combined with a domain knowledge base built upon guidelines and consensus in the field of health assessment and management for the elderly, all effective features related to functional health that can be included are preliminarily determined, including: 1) Basic information: gender, age, and death information; 2) Physical examination information: height, weight, waist circumference, systolic blood pressure, and diastolic blood pressure; 3) Dietary nutrition information: main types of staple foods, frequency of consumption of fresh fruits, frequency of consumption of fresh vegetables, types of commonly used oils, frequency of consumption of meat, frequency of consumption of fish, frequency of consumption of eggs, frequency of consumption of soy products, frequency of consumption of pickled vegetables or kimchi, frequency of consumption of white sugar or candy, frequency of consumption of garlic, frequency of consumption of dairy products, frequency of consumption of nuts, frequency of consumption of fungi and algae, frequency of consumption of vitamins or health supplements, frequency of consumption of medicinal plants, frequency of tea consumption, and the main types of tea consumed; 4) Cognitive abilities - time and space orientation: Can they identify the current time, know the Chinese zodiac animal of the year, know the date of the Mid-Autumn Festival, identify the current season, and describe their location; 5) Cognitive ability - memory: Can the first word be repeated? Can the second word be repeated? Can the third word be repeated? Can all three words be repeated at once? Can the first word be repeated after a period of time? Can the second word be repeated after a period of time? And can the third word be repeated after a period of time? 6) Cognitive ability - communication, comprehension and self-coordination ability: Can you identify and name a "pen", can you identify and name a "watch", can you repeat a sentence, can you pick up a piece of paper with your right hand, and can you fold a piece of paper with both hands; 7) Daily Living Abilities Information: Does the individual require assistance with bathing, dressing, toileting, incontinence, indoor movement, feeding, visiting neighbors independently, shopping independently, preparing food independently, and washing clothes independently; 8) Physical fitness information: Do you exercise regularly now? Did you exercise regularly in the past? Do you often do physical labor? Do you do housework? Do you grow vegetables or do other field work? Do you engage in gardening activities? Can you walk 1 kilometer? Can you lift a 5-kilogram weight? Can you squat 3 times in a row? Can you touch the base of your neck with both hands? Can you touch your lower back with both hands? Can you serve wine with both hands? Can you stand up independently while sitting in a chair? Can you pick up a book from the ground? And how many steps are required to rotate one full circle? 9) Information on depressive symptoms: Do you feel nervous or afraid? Do you feel lonely? Are you able to make your own decisions? Do you find it difficult to do things as you get older? Do you feel as happy as you did when you were young? Have you felt sad and depressed in the past two weeks? Have you lost interest in activities that usually make you happy, such as hobbies and work? And how many days have these feelings lasted? 10) Oral health: Do you have dentures? Have you had more than one toothache in the past six months? Have you had more than one buccal or jaw pain in the past six months? 11) Hearing loss information: Is there any hearing loss? 12) Disease information: Have you felt unwell in the past two weeks? Do you have high blood pressure? Do you have diabetes?

[0033] Steps 1-2: Transform and structure the above effective feature matrix into the original dataset. ,in, For the sample size, Indicate the first 1 sample feature value, where For the sample characteristic number, represent corresponding eigenvalues ∈{0,1} represents the health outcome corresponding to the sample, that is, the death state corresponding to the sample, and establishes the initial features.

[0034] Steps 1-3: Perform missing value imputation and variable type conversion on the initial features to obtain the feature set; Missing value imputation: Strict data quality control standards are employed to remove samples with a missing value ratio exceeding 50%. For the remaining samples, missing values ​​are imputed using the K-nearest neighbor algorithm, as shown in the following formula: , In the formula, Indicates the first number after filling The first sample 1 eigenvalue, This represents the number of neighbors in the K-nearest neighbors algorithm. Representation and Sample The closest in Euclidean distance A set of sample indexes, express Neighbor samples in Indicates the first The first sample There are 1 eigenvalues; where the Euclidean distance is calculated using the following formula: , In the formula, Indicates sample and samples The Euclidean distance between them Indicates the first Feature values ​​of each sample Indicates the first Feature values ​​of each sample The number of sample features.

[0035] Variable type transformation: For continuous variables in the original dataset, calculate the Youden index based on the Receiver Operating Characteristic (ROC) curve. Determine the optimal cutoff point for continuous variables. As shown in the formula below: , in, In the ROC curve, These are variables that represent all possible cutoff points and are used to divide continuous variables into two categories. Continuous variables include age, height, weight, and waist circumference.

[0036] Next, the continuous variable is represented in binary form as follows: ,in, Indicates the first The first sample Binarized eigenvalues, Indicates the first The first sample The original feature values ​​of each feature. Indicates the first The optimal cutoff point for each feature, after further processing, yields the complete feature matrix. ,in, Indicates the number of samples. The number of sample features.

[0037] Step 2: Use a machine learning model to model the mapping relationship between the preprocessed feature set and the labels, and construct a total loss function to train the machine learning model to obtain the trained machine learning model. Select key factors of functional health to construct a new set of important features.

[0038] Step 2-1: In the embodiment, the feature matrix is ​​used as the basis for the design. For input, For tags, Indicates the first For the health outcome labels of each sample, since the input feature set contains a large number of categorical variables and there may be complex nonlinear relationships between the variables, an extreme gradient boosting (XGBoost) model is used for modeling. The fitting function is... , Indicates the first Predicted health outcomes for a sample Indicates the first Features of each sample Indicates the first A decision tree, This represents the total number of trees, where each tree... Provide the predicted values ​​for the leaf nodes.

[0039] Based on the XGBoost model, the total loss function consists of a loss term and a regularization term, calculated as follows: , , In the formula, Represents the total loss function. This represents the binary cross-entropy loss term. Indicates the first Health outcome labels for each sample Indicates the first Predicted health outcomes for a sample Indicates the number of samples. Represents the regularization term, Indicates the first A decision tree, This represents the total number of trees. The regularization term representing the number of single trees, This represents the number of leaf nodes in the tree. Indicates the leaf splitting threshold. Represents the regularization coefficient. Indicates the first The weight of each leaf node.

[0040] Step 2-2: To enhance the model's handling of categorical variables, some discrete features with multiple values ​​(such as dietary habits, education level, sleep quality, etc.) are not suitable for direct one-hot encoding. Therefore, a target encoding strategy is adopted to map the categorical variables of discrete features to continuous variables based on their statistical relationship with the label y, specifically: Let the discrete feature be the first The sample value is ,in Represents the set of all possible categories for this feature; label ∈{0,1}, for each category Calculate its target encoding value : , In the formula, For indicator functions, The positive rate of the overall sample. It is a smoothing parameter. Used to prevent overfitting of classes with small sample sizes. Indicates the number of samples. The index variable represents the sample.

[0041] Finally, the first Feature values ​​of each sample Encoded as: ,in, Indicates the first The first sample The target encoding of each feature.

[0042] Steps 2-3: To fully utilize the data to validate the model's robustness, the dataset is divided into training and test sets. A five-fold cross-validation method is used to evaluate model performance, with the F1 score as the primary evaluation metric. , Meanwhile, using the F1 score as the optimization objective, a grid search is performed on the hyperparameter space of the XGBoost model to find the optimal hyperparameters to improve the performance of the XGBoost model. The hyperparameters to be searched and optimized include: learning rate, maximum tree depth, subsampling ratio, feature sampling ratio, and regularization parameter λ, to obtain the trained machine learning model.

[0043] Steps 2-4: Due to the large number of all valid features related to functional health, overfitting is likely to occur. Therefore, the weighted minimum redundancy maximum relevance (mRMR) algorithm is used to filter the preprocessed feature set. The importance weights of the XGBoost model are introduced into the mRMR algorithm, which, while preserving redundancy removal and relevance balance, tends to retain features that contribute significantly to the model, enhancing the targeting and effectiveness of feature selection; the preprocessed feature set... ,in It is the total number of features, and the target variable is... Weight vector Feature importance derived from the XGBoost model, and normalized to ensure... : , in, Indicates the first Normalized weights of each feature, Representation of features Gain value in the XGBoost model Represents the feature set The first in One characteristic, Indicates all features From 1 to The gain value is taken as the maximum value.

[0044] The goal is to extract features from the set of features. Select a subset It includes the target variable Features with the highest correlation, lowest redundancy among features, and high contribution to the model. Specifically: 1) Calculate the weighted mutual information between features and the target variable: Based on the preprocessed feature set, for each feature Calculate its relationship with the target variable Mutual information between Use the evaluation of each feature With target variable Correlation between them: , in, It is a feature and target variable The joint probability distribution, and These are features and target variable The marginal probability distribution; This leads to a weighted correlation index used to measure the relationship between features and the target: .

[0045] 2) Initialize feature subset : Initially, For an empty set, select the one with the highest weighted mutual information (relative to the target variable). Features with the highest correlation As a subset of features The first feature, Represented as: The features with the highest mutual information are selected from the preprocessed feature set. Removed from the middle.

[0046] 3) Iterative feature selection: Among the remaining features, for each feature Calculate features With target variable Maximum weighted correlation with the selected feature subset The difference in minimum redundancy is used as the candidate feature score. Select the feature with the highest score and remove it from the preprocessed feature set. Step-by-step feature selection satisfy: , in For the selected feature set, Representation of features and characteristics Weighted mutual information between them; Iteratively select features until a subset of features is reached. Includes predefined Based on these features, key factors for functional health are selected to construct a new set of important features. .

[0047] like Figure 2 As shown, key factors for functional health were screened to construct a new set of important characteristics, including: age, gender, frequency of consumption of white sugar or candy, whether one knows the zodiac animal of the year, ability to visit neighbors alone, and ability to lift 5 kg.

[0048] Step 3: Establish a multivariate logistic regression model for the new set of important features to output the advantage ratio of each important feature. Convert the advantage ratio of each important feature into a questionnaire score and define the scoring rules. Design a functional health index assessment questionnaire for the elderly population, set a total score screening threshold for the questionnaire, and classify the risk population.

[0049] Step 3-1: In this embodiment, based on the seven key features selected... Constructing a multivariate logistic regression: , In the formula, Indicates the number of important features. Indicates the first The first sample One important eigenvalue, Indicates the regression intercept. Represents the regression coefficient, corresponding to the th The log-odds ratio (log-OR) of each feature. Indicates that under a given feature Under the conditions, the first The health outcome for one sample was death ( The probability of ), Indicates a healthy outcome label; Step 3-2: Output the odds ratio (OR) of each important feature based on the established multivariate logistic regression: It is used to measure the strength of the association between a characteristic and a health outcome (such as death).

[0050] To ensure a consistent approach in the functional health index assessment questionnaire for the elderly population—that is, high scores on questions corresponding to each important characteristic represent high health risk—the OR values ​​for each characteristic are converted as follows: for Characteristics of >1: "Yes" answers are... 1 point is awarded for "No"; for The characteristic of <1: "No" answers receive 1 / 1 point for "Yes".

[0051] Next, based on the scoring rules defined above, the calculation formula is as follows: , The total score of the questionnaire is ,in, Indicates the first The advantage ratio of each feature Indicates the first The sample at the th Risk score on each feature Indicates the first The first sample One important eigenvalue.

[0052] Step 3-3: Set the total questionnaire score screening threshold and classify the high-risk population. This includes plotting the ROC curve on the validation set and calculating the Youden index. Determine the optimal cutoff point: within the range of questionnaire scores, the total questionnaire score must be greater than or equal to the Youden index. The population is classified as a high-risk group for functional health issues.

[0053] On the other hand, the present invention also provides a functional health index assessment system for the elderly, such as... Figure 3 As shown, it includes: The feature selection and preprocessing module is used to obtain cohort data from the long-term survey of health influencing factors of the elderly population, combine it with the knowledge base in the field of health assessment and management of the elderly, establish initial features, and obtain a feature set based on the initial features and preprocessing. The functional health status assessment model construction module is used to model the mapping relationship between the preprocessed feature set and the label using a machine learning model, and to construct a total loss function to train the machine learning model, thereby obtaining the trained machine learning model and screening out key factors of functional health to construct a new set of important features. The Functional Health Status Assessment and Questionnaire Design Module is used to establish a multivariate logistic regression based on a new set of important features to output the advantage ratio of each important feature, convert the advantage ratio of each important feature into a questionnaire score and define the scoring rules, design a functional health index assessment questionnaire for the elderly, set a total score screening threshold for the questionnaire, and classify high-risk groups.

[0054] It should be noted that the method for assessing the functional health index of the elderly population provided in the above embodiments and the embodiment of the system for assessing the functional health index of the elderly population belong to the same concept. For details of its implementation process, please refer to the embodiment of the method for assessing the functional health index of the elderly population, which will not be repeated here.

[0055] In summary, the functional health index assessment method and system for the elderly population provided by this invention incorporates basic information, physical examination information, dietary nutrition information, cognitive ability information, daily living ability information, physical fitness information, depressive symptoms information, oral health information, hearing loss information, and disease information to establish a feature set that comprehensively describes the functional health status of the elderly population. Using this comprehensive feature set as input, a basic machine learning model is trained to construct a functional health index assessment model adapted to the elderly population. Key functional health factors are selected and designed into a simple questionnaire, enabling rapid and efficient assessment of the functional health of the elderly population, and is suitable for promotion and use in communities.

[0056] Furthermore, the terms "upper," "lower," "inner," "outer," "front," and "rear" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of components and steps described in these embodiments do not limit the scope of the invention. Of course, the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of the invention. All equivalent changes or modifications made to the structures, features, and principles described in the claims of this invention should be included within the scope of the claims of this invention.

[0057] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for assessing the functional health index of the elderly population, characterized in that, Includes the following steps: Step 1: Obtain cohort data from the long-term study on health influencing factors in the elderly population, combine it with the knowledge base in the field of health assessment and management of the elderly, determine the effective features, and preprocess the effective features to obtain the feature set; Step 2: Use a machine learning model to model the mapping relationship between the preprocessed feature set and the labels, and construct a total loss function to train the machine learning model to obtain the trained machine learning model. Select key factors of functional health to construct a new set of important features. Step 3: Based on the new set of important features, establish a multivariate logistic regression to output the advantage ratio of each important feature. Convert the advantage ratio of each important feature into a questionnaire score and define the scoring rules. Design a functional health index assessment questionnaire for the elderly population, set a total score screening threshold for the questionnaire, and classify the risk population.

2. The method for assessing the functional health index of the elderly population according to claim 1, characterized in that, In step 1, the effective features are preprocessed to obtain a feature set, including: structuring the effective features into the original dataset, and performing missing value imputation and variable type conversion on the original dataset to obtain the feature set; The variable type conversion includes: for continuous variables in the original dataset, calculating the Youden index using ROC curves. Determining the optimal cutoff point for continuous variables And binarize the continuous variables to output the feature set. ,in , In the formula, Indicates the first The first sample Binarized eigenvalues, Indicates the first The first sample The original feature values ​​of each feature. Indicates the first The optimal cutoff point for each feature Indicates the number of samples. The number of sample features.

3. The method for assessing the functional health index of the elderly population according to claim 2, characterized in that, Continuous variables include: age, height, weight, and waist circumference.

4. The method for assessing the functional health index of the elderly population according to claim 1, characterized in that, When using a machine learning model to model the mapping relationship between the preprocessed feature set and the label, a target encoding strategy is used to encode the preprocessed feature set. Among them, discrete features with multiple values ​​are encoded using the target encoding strategy, which maps the categorical variables of discrete features to continuous variables according to their statistical relationship with the label. The discrete features include dietary habits, education level, and sleep quality.

5. The method for assessing the functional health index of the elderly population according to claim 1, characterized in that, In step 2, the machine learning model is the XGBoost model.

6. The method for assessing the functional health index of the elderly population according to claim 5, characterized in that, In step 2, the constructed total loss function includes a loss term and a regularization term, and the calculation formula is as follows: ， ， In the formula, Represents the total loss function. Indicates the loss item. Indicates the first Health outcome labels for each sample Indicates the first Predicted health outcomes for a sample Indicates the number of samples. Represents the regularization term. Indicates the first A decision tree, This represents the total number of trees. The regularization term representing the number of single trees, This represents the number of leaf nodes in the tree. Indicates the leaf splitting threshold. Represents the regularization coefficient. Indicates the first The weight of each leaf node.

7. The method for assessing the functional health index of the elderly population according to claim 1, characterized in that, In step 2, the process of screening out key factors of functional health to construct a new set of important features includes: based on the preprocessed feature set, introducing XGBoost importance weights in the minimum redundancy maximum relevance algorithm framework to perform feature screening, and obtaining a new set of important features including age, gender, frequency of consumption of white sugar or candy, whether one knows the zodiac animal of the year, whether one can visit neighbors alone, and whether one can lift a 5 kg weight.

8. The method for assessing the functional health index of the elderly population according to claim 1, characterized in that, Step 3, which involves converting the odds ratios of each important feature into questionnaire scores and defining scoring rules, includes: For important features with an odds ratio greater than 1, the score for a yes answer is the odds ratio, and the score for a no answer is 1. For important features where the odds ratio is less than 1, the score for a yes answer is 1 / odds ratio, and the score for a no answer is 1. Based on the scoring rules defined above, the calculation formula is as follows: ， The total score of the questionnaire is ,in, Indicates the first The advantage ratio of each feature Indicates the first The sample at the th Risk score on each feature Indicates the first The first sample One important eigenvalue.

9. The method for assessing the functional health index of the elderly population according to claim 1, characterized in that, In step 3, the setting of the total questionnaire score screening threshold and the classification of risk groups includes: calculating the Youden coefficient, and classifying people whose total questionnaire score is greater than or equal to the Youden coefficient as risk groups within the questionnaire score range.

10. A functional health index assessment system for the elderly, characterized in that, The elderly population functional health index assessment system uses the elderly population functional health index assessment method according to any one of claims 1 to 9, including: The feature selection and preprocessing module is used to obtain cohort data from the long-term survey of health influencing factors of the elderly population, combine it with the knowledge base in the field of health assessment and management of the elderly, establish initial features, and obtain a feature set based on the initial features and preprocessing. The functional health status assessment model construction module is used to model the mapping relationship between the preprocessed feature set and the label using a machine learning model, and to construct a total loss function to train the machine learning model, thereby obtaining the trained machine learning model and screening out key factors of functional health to construct a new set of important features. The Functional Health Status Assessment and Questionnaire Design Module is used to establish a multivariate logistic regression to output the advantage ratio of each important feature from a new set of important features, convert the advantage ratio of each important feature into a questionnaire score and define the scoring rules, design a functional health index assessment questionnaire for the elderly, set a total score screening threshold for the questionnaire, and classify high-risk groups.

Images